Self-supervision in Audio, Speech and Beyond

Accepted papers will be available throughout the IEEE Xplore Digital Library and appear in the conference proceedings.

Accepted papers will therefore be archived and transferred to IEEE.

The workshop welcomes the submission of work on, but not limited to, the following research directions:

1. New self-supervised proxy tasks or new approaches using self-supervised models in speech and audio processing. 

2. Theoretical or empirical studies focusing on understanding why self-supervision methods work for speech and audio. 

3. Exploring the limits of self-supervised learning approaches for speech and audio processing, for example, adverse environment conditions, multiple languages, or generalization across downstream tasks. 

4. Comparison or integration of self-supervised learning methods and transfer learning methods in speech and audio processing tasks. 

5. Self-supervised learning approaches involving the interaction of speech/audio and other modalities. 

6. New evaluation protocols and benchmarks for SSL models within the speech and audio modalities. 

7. Studies underlining and/or solving biases and robustness issues with SSL models. 

8. Novel approches to increase the compute and data efficiency of SSL algorithms. 

9. Papers describing newly trained SSL models covering new languages and/or modalities, accompanied by proper analysis.